Mentally Spent: Credit Market Conditions and Mental Health

Qing Hu, Ross Levine, Chen Lin, and Mingzhu Tai∗

March 2019

Abstract

In light of the human suffering and economic costs associated with mental illness, we provide the

first assessment of whether local credit conditions shape the incidence of mental depression.

Using several empirical strategies, we discover that bank regulatory reforms that improved local

credit conditions reduced mental depression among low-income households and the impact was

largest in counties dominated by bank-dependent firms. On the mechanisms, we find that the

regulatory reforms boosted employment, income, and mental health among low-income

individuals in bank-dependent counties, but the regulatory reforms did not increase borrowing by

these individuals.

JEL Codes: I1; G21; D14; R23

Keywords: Health; Financial Institutions; Personal Finance; Regional Labor Markets

∗ Hu, Faculty of Business and Economics, The University of Hong Kong: hqhuqing@connect.hku.hk; ; Levine, Haas School of Business, University of California, Berkeley: rosslevine@berkeley.edu; Lin, Faculty of Business and Economics, The University of Hong Kong; chenlin1@hku.hk; Tai, Faculty of Business and Economics, The University of Hong Kong: taimzh@hku.hk.

1

1. Introduction

Almost one in five U.S. adults experienced mental illness in 2016 and over three million

became recipients of Social Security and Disability Insurance benefits due to a mental health

condition. Insel (2008) estimates that mental illness lowers U.S. earnings by $193 billion per

annum due to absenteeism and lower productivity. Looking globally, the World Health

Organization (2011) reports that mental illness is the leading cause of lost working hours and

estimates that the global cost of mental illness in 2010 was nearly $2.5 trillion.1

In light of the human suffering and economic costs from mental illness, we evaluate the

impact of financial systems on mental health. Although extensive research shows that the

competitiveness and efficiency of credit markets shape economic welfare by influencing

economic growth (King and Levine 1993; Jayaratne and Strahan 1996; Levine and Zervos 1998;

Rajan and Zingales 1998), income inequality (Beck, Levine, and Levkov 2010), entrepreneurship

(Black and Strahan 2002; Kerr and Nanda 2009; Levine and Rubinstein 2017, 2018), household

consumption and borrowing (Agarwal and Qian 2014; Agarwal et al 2018), and access to

education and homes (Jacoby 1994; Sun and Yannelis 2016; Favara and Imbs 2015), we are

unaware of any previous research on the impact of financial systems on mental health.

Existing research provides differing perspectives on whether better functioning financial

systems improve or worsen mental health. On the one hand, research shows that (1) easing credit

constraints boosts economic conditions, especially among low-income households, as

summarized in Popov (2018), and (2) adverse economic conditions—such as low income,

unemployment, and job insecurity— are positively associated with mental illness, as shown by

Lund et al (2010 2011), Reeves et al (2012), and Apouey and Clark (2015), Wickham et al

(2017), and Adhvaryu, Fenske, and Nyshadham (2019). These findings, therefore, suggest both

that easing credit constraints will reduce mental illness and the impact on mental illness is likely

to most pronounced among lower-income individuals, who suffer from higher rates of mental 1 On U.S. Social Security and Disability Insurance, see https://www.nimh.nih.gov/health/statistics/mental-illness.shtml and the Annual Statistical Report on the Social Security Disability Insurance Program, 2016. For information on the costs of mental illness, see Bartel and Taubman (1986), Kessler (2012), World Health Organization (2015, 2018), and Chisholm et al (2016).

2

illness and who experience larger boosts in income and employment when credit conditions ease

(e.g., Beck, Levkov, and Levine (2010). On the other hand, there are potentially countervailing

influences. For example, Kerr and Nanda (2009) find that improvements in credit conditions

increase the rates of business entry and exit. While spurring economic dynamism, more firm

entry and exist might also increase anxiety about job stability, with adverse effects on mental

health. Furthermore, to the extent that easing credit constraints induces a surge in household

indebtedness, this too can intensify household stress and harm mental health, as suggested by

Clayton, Linares-Zegarra, and Wilson (2015) and Berger, Collins, and Cuesta (2016).

To evaluate the impact of credit market conditions on mental health and shed empirical

light on the mechanisms linking finance and mental health, we exploit an exogenous source of

variation in local credit conditions: the deregulation of interstate bank branching that began in

1995 and continued through 2005 (e.g., Johnson and Rice 2008). Although the Riegle-Neal Act

eliminated regulatory prohibitions on interstate banking in 1995, it did not prohibit states from

erecting barriers to out-of-state branch expansion. Since the costs of branching are lower than the

costs of establishing subsidiaries, interstate branching restrictions limited competition between

banks in different states and made banking markets less efficient. Rice and Strahan (2010) show

that lowering regulatory barriers to interstate branching improved credit conditions by reducing

the interest rates charged by banks on loans to small firms. Building on their strategy, we use

cross-state, cross-time variation in the removal of regulatory impediments to interstate bank

branching as an exogenous source of variation in local credit conditions and examine the impact

of these regulatory reforms on mental health.

We explore the impact of these bank regulatory reforms on mental illness using data from

the National Longitudinal Survey of Youth 1979 (NLSY79). The NLSY79 is a nationally

representative survey that follows individuals born in the years 1957-1964. There are 12,686

individuals, who were between the ages of 14 and 22 when the survey started in 1979, and the

NLSY79 continues to survey these individuals. In addition to information on employment,

income, education, wealth, demographics, cognitive and noncognitive traits, and family

3

background, the NLSY79 also obtains information on each individual’s physical and mental

health. For mental health, the NLSY79 asks seven questions that comprise the Center for

Epidemiological Studies (CES) depression index. We examine (a) the answers to each of the

seven questions separately, (b) the composite depression index based on equally weighting the

answers to the seven questions, and (c) several other measures of depression based on different

weighting schemes. We focus our discussion on the results using the composite, equally

weighted depression index. Since the NLSY79 provides these mental health measures over time

for each individual, we test whether changes in credit conditions trigger changes in the mental

health of individuals.

We use three empirical strategies to assess the impact of interstate branch deregulation on

mental health. First, exploiting the cross-state, cross-time variation in bank deregulation and the

longitudinal nature of the NLSY79, we examine the response of individual mental health to

branch deregulation while conditioning on individual fixed effects, gender-race-year fixed

effects, state-specific linear time trends, and time-varying state characteristics, such real gross

state product (GSP) per capita and population. Furthermore, we also differentiate between high-

and low-income individuals. As noted above, bank deregulation exerts an especially large impact

on the earning of lower income individuals, so we test whether the relationship between branch

deregulation and mental health is more pronounced among lower-income individuals. By

controlling for individual and gender-race-year effects, we condition out time-invariant

individual-level characteristics that shape mental health and time-varying factors that

differentially affect these sub-population groups.

The second empirical strategy differentiates among counties within states and uses a

spatial regression discontinuity research design to compare neighboring counties that are subject

to different interstate branching policies. Huang (2008) compared counties across state borders

when assessing the impact of interstate banking deregulation that allowed banks to establish

subsidiaries in another state. We apply this method to interstate branch deregulation that allowed

banks to establish branches in other states. Thus, we compare two otherwise similar counties

4

separated by a state border in which the states have different interstate branching regulations and

evaluate the impact of changes in these branching regulations on changes in mental health in the

counties. If the results from this spatial regression discontinuity strategy confirm the earlier

findings, this will reduce concerns that omitted variables drive the findings on the relationship

between finance and mental health.

Our third empirical strategy also differentiates among counties within states and explores

two potential mechanisms through which credit conditions shape mental health: The “firm credit”

and “consumer credit” mechanism. The firm credit mechanism holds that enhancing credit

conditions will improve mental health by boosting income and employment. We evaluate this

potential mechanism in two ways. First, we differentiate counties by the degree to which the

county’s firms depend on external finance, namely banks. If easing credit constraints improve

mental health by boosting income and employment, then we should find a particularly large

impact of deregulation on mental health in counties where bank branch deregulation is likely to

have the biggest effect on firm credit conditions: counties with firms that depend heavily on

banks. Following Rajan and Zingales (1998), we differentiate counties by the degree to which

firms are financially dependent and examine whether interstate branch deregulation exerts an

especially pronounced effect on mental health among low-income workers and in counties

dominated by financially-dependent firms. Second, we differentiate counties by the density of

bank branches before bank branch deregulation. Holding other features of the counties constant,

we expect that bank branch deregulation will have a bigger effect on firms’ access to credit in

“under-banked” counties than in counties that were already densely banked before deregulation.

The second mechanism—the consumer credit mechanism—holds that enhancing credit

conditions will improve mental health by easing household access to credit, without necessarily

increasing income or employment. To assess this consumer credit, we test whether interstate

bank branch deregulation triggered an increase in borrowing among lower-income workers.

We first discover that (1) interstate branch deregulation is associated with a material drop

in mental depression, (2) this effect is driven by improvements in mental health among low-

5

income individuals, and (3) these results hold when implementing the spatial regression

discontinuity strategy. These findings are robust to including individual and gender-race-year

fixed effects, state-specific linear time trends, and time-varying state characteristics. The

estimated impact is large. If the average U.S. state relaxes interstate branching restrictions by one

degree—where branch restrictions range from zero (most restrictive) to four (least restrictive),

the estimates indicate that the composite depression index (Depression Index) for the low-

income group drops by 0.522, which represents a 15% drop relative to the sample mean, and

clinical depression falls by 3.8%.

Turning next to the mechanisms through which credit conditions shape mental health, we

first find that interstate branch deregulation triggers an especially pronounced increase in income

and employment and decrease in mental depression (a) among low-income workers within

counties dominated by bank-dependent firms and (b) among low-income workers within

counties with low pre-deregulation bank density. Thus, following branch deregulation, the same

group of individuals experienced both a disproportionately large increase in employment and

income and especially large decrease in mental depression. These findings are consistent with the

firm-credit channel: easing firm credit constraints boosts income and employment and these

improvements enhance mental health. By focusing on a group of individuals most likely to be

affected by interstate branch deregulation—low-income workers in bank-dependent counties, we

reduce concerns that omitted variables confound our findings. Second, we do not find evidence

consistent with the consumer-credit channel. In particular, we find no evidence that interstate

branch deregulation increases borrowing by individuals, with beneficial—or adverse—effects on

mental depression.

Our work relates to, but is distinct from, three influential lines of research. First, a

growing body of work shows that the economic conditions facing parents shape the mental

health of their children. In a study of households in Ghana, Adhvaryu, Fenske, and Nyshadham

(2019) shows that children in families that experience adverse economic shocks have higher

incidence of mental disease later in life. Focusing on a different form parental stress, Persson and

6

Rossin-Slater (2018) finds that prenatal exposure to the death of a maternal relative increases

mental illness later in life in a longitudinal study of Swedish families. In our study, we show that

bank-deregulation induced changes in the economic conditions facing individuals materially

shapes their mental health, which could tied to the mental health of their children as research

shows a strong link between the mental health of parents and the cognitive and noncognitive

development of their children (e.g., Ramchandani et al 2005). Second, an important series of

papers explores the impact of recession on health. The studies by Ruhm (2000, 2005, and 2015)

and Miller et al (2009) show that recessions are typically associated with improvements in

physical, but not mental, health (Gili et al 2013). Although our research also assesses the

connection between economic conditions and health, we focus on a regulatory change that

improves the functioning of local credit markets and we then trace the effects through to mental

health in those locals. Finally, our work relates to a growing body of research on how the

financial system influences household welfare. While an established line of research studies the

impact of finance on national, industrial, and firm growth, this newer line of research focuses on

the relationship between finance and income inequality (Beck, Levine, and Levkov 2010),

gender and racial discrimination (Black and Strahan 2001; Levine, Levkov, and Rubinstein

2014), crime (Garmaise and Moskowitz 2006), and educational opportunities (Popov 2014; Sun

and Yannelis 2016). We contribute to this work by examining the impact of credit conditions on

mental health, which is a key component of welfare.

The remainder of the paper proceeds as follows. Section 2 discusses the Riegle-Neal Act

and the details of interstate branch deregulation, while Section 3 presents the data and variable

definitions. Section 4 presents the baseline result and spatial regression discontinuity design.

Section 5 explores potential channels and deepens understanding of factors influencing mental

health condition and subjective well-being. Section 6 concludes.

7

2. Bank Branch Deregulation

The Riegle-Neal Interstate Banking and Branching Act (IBBEA) of 1994 effectively

removed (a) restrictions on banks expanding across state lines through the establishment of

separately capitalized bank subsidiaries and (b) federal barriers to interstate branching. As

explained by Johnson and Rice (2008), however, the Act provided states with the discretion to

limit bank expansion across state lines through the establishment of bank branches. Interstate

branching can occur through two means: (a) an out-of-state bank can acquire an in-state bank

and convert that bank into its branches or (b) an out-of-state bank can either establish new

branches within a state (“de novo” branching) or purchase the branches of an in-state bank.

Following Riegle-Neal, many states erected barriers to banks from other states

establishing branches within their borders. States used four types of regulatory restrictions on

interstate branching. First, some states imposed minimum age restrictions with respect to how

long a target bank has been in existence before it can be acquired and consolidated into branches.

These minimum age restrictions, which have a federally set maximum of five years, make cross-

state banking more costly because they required that banks (a) purchase an entire older bank,

which is more costly than opening a branch, or (b) open a new subsidiary and then wait until the

minimum age restriction has been satisfied before converting the subsidiary into a branch.

Second, some states prohibited de novo interstate branching. Third, some states prohibited the

acquisition of a single branch or portions of an institution, which represents an additional barrier

to cross-state branching. Fourth, some states imposed limits on the percentage of insured

deposits in state that a single bank could hold, which could limit large interstate bank mergers.

Building on Johnson and Rice (2008), Rice and Strahan (2010) construct an index of

regulatory restrictions on interstate branching. Based on their data, our IBBEA Index takes a

value between zero and four, where we add one point to the index if the state (1) does not impose

a minimum age restriction for acquisition, (2) allows de novo interstate branching, (3) permits

interstate branching by acquiring a single branch; (4) sets the deposit-cap no less than 30%. Thus,

8

larger values indicate a more deregulated interstate branching environment. We use this index to

exploit post-1994, cross-state variations in restrictions on interstate branching.

3. Data and Variables

3.1 Individual-Level Data

We extract data from the National Longitudinal Surveys Youth 1979 (NLSY79)

administered by the U.S. Bureau of Labor Statistics. The NLSY79 is a nationally representative

survey that follows a sample of American youth born in the years 1957-1964. A total of 12,686

individuals (aged 14 -22 years) were interviewed in the initial survey year, 1979. The surveys

were conducted annually from 1979 to 1994, and then biannually afterwards. Following Altonji

and Pierret (2001) and Levine and Rubinstein (2017), we restrict the NLSY79 sample as follows.

We drop individual-year observations with a missing state identifier and those not residing in one

of the 50 states or the District of Columbia. We also drop observations with missing data on key

variables and those who change their residence state over the sample period. We use the sample

weights provided by NLSY79.

The NLSY79 has two advantages over several other individual-based U.S. datasets. First,

it is a representative, longitudinal survey. The long-panel nature of the NLSY79 allows us to

control for individual fixed effects to better identify the impact of interstate branch deregulation

on mental health. Second, in addition to containing information on standard demographic and

economic traits, the NLSY79 also contains information on family background, physical and

mental health, and an assortment of other cognitive and noncognitive traits.

In our analyses, we control for several time-varying, state-specific traits that are not part

of the NLSY79 data. Gross State Product (GSP) per capita and total population come from the

Bureau of Economic Analysis. We adjust all nominal values into 2010 prices using CPI prices.

9

3.2 Key Variables

3.2.1 Depression Measures

We use measures of each respondent’s level of depression based on questions in the

Attitudes, Health, and Health Module 40 & Over sections of the NLSY79. The NLSY79 surveys

each individual’s mental health in 1992, 1994, and the year immediately after the person

becomes 40 years old (over 1998-2005). Thus, we have three measures of each individual’s

mental health. For most of the interviewees, therefore, we have two depression measures before

the 1994 Riegle-Neal Act and one measure afterwards. Given data on mental health and the

index of restrictions on interstate bank branching, our sample period runs from 1992 through

2005.

We utilize the Center for Epidemiologic Studies Depression Scale, which measures

symptoms of depression using seven sets of questions. Respondents were asked to evaluate their

degree of agreement (during the past week) with the following statements:

(1) I did not feel like eating; my appetite was poor (Trouble Eating),

(2) I had trouble keeping my mind on what I was doing (Trouble Concentrating),

(3) I feel depressed (Feel Depressed),

(4) I felt that everything I did was an effort (Everything an Effort),

(5) My sleep was restless (Trouble sleeping),

(6) I felt sad (Feel Sad), and

(7) I could not get “going” (Lethargic).

For each statement, respondents were asked to choose from (a) Rarely/None of the time/1 Day,

(b) Some/A little of the time/1-2 Days, (c) Occasionally/Moderate amount of the time/3-4 Days,

and (d) Most/All of the times/5-7 Days. Corresponding ordinal values of 0 to 3 points are

assigned to these four responses.

Besides ratings on the seven individual items, NLSY79 also reports a total score

(Depression Index) that aggregates the individual responses to a single number spanning from 0

(individual chooses 0 as an answer for all seven questions, least depressed) to 21 (individual

10

chooses 3 for all dimensions, most depressed). This Depression Index is often employed in

epidemiological and treatment studies to screen for depression disorder in the community, and a

cutoff value of 8 is widely used to detect clinical depression (Levine, 2013).

To measure whether an individual passes a particular threshold measure of “clinical

depression,” we create dummy variables based on the Depression Index. Specifically, if the

Depression Index is above 8, then D-Depression Index Above 8 equals one and equals zero

otherwise; if the Depression Index is above 10, then D-Depression Index Above 10 equals one

and equals zero otherwise; and if the Depression Index is above 12, then D-Depression Index

Above 12 equals one and equals zero otherwise. Thus, in addition to the Depression Index, we

examine these three indicators of severe depression.

As an additional robustness test, we create and examine measures based on the intensity

of the answers to the seven questions regarding mental depression. In particular, % of Answers

“0” equals the percentage of the seven questions that the respondent answers, “Rarely/None of

the time/1 Day per week;” % of Answers “1” equals the percentage of the seven questions that

the respondent answers, “Some/A little of the time/1-2 Days per week;” and % of Answers “2”

and % of Answers “3” are defined analogously.

3.2.2 Individual-Level Characteristics

Inspired by a growing literature on the “economics of happiness” and subjective well-

being (Dolan, Peasgood and White, 2008), we examine several individual-level characteristics.

With respect to basic traits, we examine Education and Age, which equal the highest grade

completed by the respondent, ranging from below high school (less than 12 years of schooling)

to advanced graduate (over 16 years of schooling), and the age of the difference between the

birth year and the survey year respectively. With respect to family background, we examine the

following. FamilyIncome7981 equals the natural logarithm of one plus the average total family

income over the 1979 to 1981 period (in 2010 dollars). MotherEducation and FatherEducation

equals the maximum years of schooling received by the respondent’s mother and father

11

respectively. D-BothParents is a dummy variable that equals one if the respondent lived with

both parents at the age 14, and zero otherwise.

To capture an individual’s aptitudes, attitudes, and personality traits, we examine the

following. AFQTAbilityTest equals the respondent’s percentile score from taking the Armed

Services Vocational Aptitude Battery (ASVAB) test in 1980, where the test is designed to

measure knowledge and skill in the areas of arithmetic reasoning, word knowledge, paragraph

comprehension and numerical operations. RosenbergSelfEsteem equals the individual’s score

using the Rosenberg Self-Esteem Scale, which is computed from respondent’s answers in 1987

to questions designed to measure self-worth (Rosenberg, 1979).2 PearlinSenseOfControl equals

the Pearlin Mastery score based on questions from the 1992 survey and is based on answers to

seven questions designed to measure the respondent’s sense of control over one’s life (Pearlin et

al 1981).3 D-RiskAttitude is a dummy variable that equals one if the respondent answers “Yes”

facing the question “Would you take a job that could either double family income or cut income

by a third” in the 1993 survey year, and zero otherwise. This dummy variable proxies for the

respondent’s level of risk tolerance just before 1994 IBBEA deregulation.

2 Respondents were asked to choose (1) Strongly Agree, (2) Agree, (3) Disagree, and (4) Strongly Disagree over ten statements, such as “I am a person of worth,” “I have a number of good qualities” etc. The total score ranges from 0 to 30, with a higher value indicating a higher level of self-esteem. 3 Respondents were asked to choose (1) Strongly Disagree, (2) Disagree, (3) Agree, and (4) Strongly Agree over seven statements, such as “I have little control over the things that happen to me,” “There is little I can do to change many of the important things in my life” etc. The total score ranges from 7 to 28, with a higher value indicating a more positive self-concept and a stronger internal control over one’s destiny.

12

3.2.3 Variables for the Channel Explorations

We also examine the impact of interstate bank branch deregulation on each individual’s

economic conditions. D-WorkHours is a dummy variable that equals one if the respondent

participates in the job market and has positive working hours in a given year, and zero otherwise.

Log(WorkHours) takes the natural logarithm of one plus total working hours. Log(EmpWeeks)

takes the natural logarithm of one plus total employment weeks. Log(TotalIncome) is the natural

logarithm of one plus the respondent’s total income (in 2010 dollars) including wages, salaries,

and income from farm or business.

Finally, we examine the degree to which individuals borrow. Leverage equals the ratio of

total liabilities (excluding mortgages on residential property, such as debts on

farm/business/other property, debts on vehicles etc.) to total assets (excluding the market value

of residential property, such as the market value of the farm/business/other property, vehicles,

etc.). D/I Ratio is the ratio of total liabilities excluding mortgages on residential property as

defined above over respondent’s total income (including wages, salaries, and income from farm

or business).

3.2.4 State-Level Controls

We also control for the following time-varying, state-specific influences. Log(GSP Per

Capita) and Log (Population) are the natural logarithm of Gross State Product (GSP) per capita

and total state population respectively and are obtained from the Bureau of Economic Analysis.

3.3 Summary statistics

Table 1 provides summary statistics on key variables using the sampling weights

provided by the NLSY79 over our sample period 1992-2005, and Appendix 1 provides detailed

variable definitions.4 The Depression Index ranges from 0 (not depressed at all) to 21 (most

depressed), and the sample average is 3.5. 13.6% of respondents suffer from clinical depression

as measured by a Depression Index score that is above eight. 4 All continuous variables are winsorized at the 1st and 99th percentiles.

13

Figure 1 graphs the average value of the Depression Index from 1992 through 2014. As

shown, depression falls during the period of rapid economic growth in the 1990s, but rises

sharply following the 2008 global financial crisis. In our analyses, we will focus on individual-

level data and will condition out such trends, but the aggregate movements are consistent with

past work showing the positive association between economic conditions and mental health.

4. Empirical Results

In this section, we first exploit the cross-state, cross-time variation in bank deregulation

and the longitudinal nature of the NLSY79 to examine the relationship between branch

deregulation and mental health while differentiating between high- and low-income individuals

and conditioning on individual, state, and gender-race-year fixed effects as well as time-varying

state economic conditions. We also show that the results are robust to several sensitivity analyses,

including the inclusion of state specific linear time trends. We then exploit the county-level data

and employ a spatial regression discontinuity research design to compare neighboring counties

that are subject to different interstate branching policies.

4.1 Baseline Results

We use a generalized difference-in-difference estimation strategy that exploits the

staggered lowering of regulatory impediments to interstate state branching to evaluate the impact

of bank regulatory reforms on mental depression. Thus, our analyses employ the following panel

regression model:

𝑌𝑖,𝑠,𝑡 = 𝛼 + 𝛽 ∙ 𝐼𝐼𝐼𝐼𝐼 𝐼𝐼𝐼𝐼𝐼𝑠,𝑡 + 𝜃 ∙ 𝐶𝐶𝐼𝐶𝐶𝐶𝐶𝐶𝑖,𝑠,𝑡−1 + 𝛿𝑖 + 𝛿𝑔,𝑟,𝑡 + 𝜀𝑖,𝑠,𝑡, (1)

where 𝑌𝑖,𝑠,𝑡 measures the mental health of individual i from state s at time t. 𝐼𝐼𝐼𝐼𝐼 𝐼𝐼𝐼𝐼𝐼𝑠,𝑡 is

the interstate bank branching deregulation index of state s at time t. 𝐼𝐼𝐼𝐼𝐼 𝐼𝐼𝐼𝐼𝐼𝑠,𝑡 ranges from

zero to four, where higher values indicate fewer restrictions on interstate branch banking. 𝛽 is

our key coefficient of interest, capturing the impact of bank branch deregulation on mental health.

14

𝐶𝐶𝐼𝐶𝐶𝐶𝐶𝐶𝑖,𝑠,𝑡−1 is a matrix of (a) time-varying state-level characteristics, including the log of

GSP per capita and the log of total population and (b) time-invariant individual-level traits,

including the income of the individual’s family when the individual was young, the education

level of each parent, whether the individual was raised in a home with both parents, the

individual’s AFQT, Rosenberg, Pearlin Mastery, and Risk Aversion scores. In some

specifications we include individual fixed effects, 𝛿𝑖, instead of these individual-level traits.

𝛿𝑔,𝑟,𝑡 denotes gender-race-year fixed effect. All regressions use the survey weights provided by

NLSY79 to adjust for non-response, clustering, and stratification issues inherent in the survey

data. Standard errors are clustered at the state level.

Table 2 presents the results from estimating equation (1) while using various

combinations of controls and using different subsamples of individuals. Columns (1) to (3) report

results with state and gender-race-year fixed effects and an extensive list of time-invariant

individual traits. Columns (4) to (6) include individual fixed effects. By including individual and

gender-race-year effects, we condition out time-invariant individual-level characteristics that

might shape mental health and time-varying factors that differentially affect these sub-population

groups. Following Chetty, Hendren, Kline and Saez (2014), we differentiate between the low-

and high-income subsamples by (1) computing average total income (in 2010 dollars), including

wages, salaries and income from farm or business, over the three-year window immediately

before our sample period, (i.e., 1989-1991) and (2) dividing each state’s sample of individuals

into low- and high income subsamples based on the state median value of average total income

over the 1989-1991 period.

As shown in columns (1) – (3), we discover that bank branch deregulation is associated

with a sharp drop in depression among lower income individuals, but not among the high-income

group. The estimated coefficient on IBBEA Index is both economically large and statistically

significant among the sample of low-income individuals. The point estimate for the low-income

group is -0.284, which implies that one more degree of interstate bank branch openness, i.e., an

increase of the IBBEA Index of one, is associated with an 8% reduction in the Depression Index

15

relative to the sample mean (0.08=0.284/3.508). In contrast, the point estimate for the high-

income group is -0.053 and statistically insignificant. The difference between the estimates for

the low- and high-income groups is statistically significant as measured by the Wald test. The

remaining coefficients in column (1) to (3) are consistent with existing research on the

“economics of happiness.” For example, individuals from higher-income families

(FamilyIncome7981) and those who were raised in two-parent households (D-BothParents) tend

to suffer from less mental depression. We also find that individuals with higher learning aptitude

scores, greater self-esteem, and a stronger sense of internal control over one’s life also have

lower depression scores.

When also conditioning on individual fixed effects, we continue to find that lowering

regulatory barriers to interstate bank branching is associated with a material improvement in

mental health among low-income individuals, but not among high-income individuals. As shown

in columns (4) - (6) of Table 2, lower regulatory barriers to interstate bank branching is

associated with a material improvement in mental health among low-income individuals. For

example, the coefficient in column (5) suggests that one more degree of interstate bank branch

openness—as measured by the IBBEA Index that ranges from 0 to 4—reduces the Depression

Index by almost 0.297, which represents an 8.5% decline of the sample mean.

We next provide graphical evidence of the validity of our difference-in-difference

methodology. Figure 2 traces the evolution of the total Depression Index for low-income

individuals after dividing those individuals into (a) those living in states that ultimately

deregulate interstate bank branch regulations more aggressively and ultimately have above

median levels of the IBBEA Index and (b) those living in states that ultimately deregulate less

aggressively and then have below median IBBEA Index levels. There are three observations for

each individual’s mental health: two before the period of interstate bank branch deregulation

started and one afterwards. We define individuals as living in high- or low- IBBEA Index states

by the median (i.e., 2) of the IBBEA Index after 1994, which corresponds to the third observation

for each person. The figure plots the average values for low-income individuals for each of these

16

three observations while splitting the sample between those living in high- and low-IBBEA Index

states. Figure 2 shows that, on average, those individuals living in states that ultimately

deregulate more aggressively and nave comparatively high IBBEA Index values experience a

much larger drop in mental distress relative to the pre-deregulation period. In particular, there is

a parallel downward trend of mental depression before the deregulation period, as depression

tends to fall in older individuals. Then there is a sharp break in the trend for those living in states

that deregulate interstate bank branch regulations more aggressively.

4.2 Baseline Results: Robustness Tests

We next focus on severe depression by examining D-Depression Index Above 8, D-

Depression Index Above 10, and D-Depression Index Above 12. These dummy variables indicate

whether the individual has a Depression Index value above 8, 10, and 12 respectively, where the

cut-off value of 8 is often used to gauge clinical depression. We use these as the dependent

variables in columns (1) to (3) of Table 3. Furthermore, as an additional robustness test, we

examine the intensity of the answers to the seven questions regarding mental depression using

the four variables defined above: % of Answers “0,” % of Answers “1,” % of Answers “2,” and %

of Answers “3.” Table 3 also provides the results on these four mental depression indicators.

As shown in Table 3, the results on severe depression are fully consistent with the earlier

results using the Depression Index: reducing regulatory restrictions on interstate branching is

associated with a material drop in the incidence of severe depression among low-income

individuals, but not among high-income individuals. For example, among the low income group,

the estimated coefficients indicate that one more degree of interstate branch deregulation (as

measured by the IBBEA Index) is associated with a drop in the probability of severe mental

depression of between 1.4% and 1.8% depending on which cutoff is employed (e.g., D-

Depression Index Above 8, D-Depression Index Above 10, or D-Depression Index Above 12. The

results in Table 3 also emphasize the negative impact of bank deregulation on depression among

low income individuals when using the alternative indicators of depression that focus on

17

intensity of the answers to the seven questions used to gauge depression (columns 4 to 7).

Following a reduction in the restrictions on interstate branch deregulation (an increase in IBBEA

Index), individuals are much more likely to answer that they rarely or never experience the seven

features of depression and are much less likely to answer that they experience the seven features

of depression 5-7 days per week.

We also examine the relationship between interstate branch regulatory reforms and

responses to each of the seven questions underlying the composite Depression Index. As reported

in Table 4, we find that regulatory reforms that improve the operation of banking systems

improve the mental well-being of low-income individuals with respect to Trouble Concentrating,

Feel Depressed, Everything an Effort, Trouble sleeping, and Feel Sad. This examination of the

seven core aspects of depression largely confirms the earlier findings that bank regulatory

reforms that improve credit conditions tend to reduce mental distress among low-income

individuals.

Furthermore, we show that the results are robust to including state-specific linear time

trends. Although the analyses above include individual (and hence state) fixed effects, there

might be trends in state conditions that account for the reduction in depression following the

relaxation of interstate branching restrictions. When we control for state-specific linear time

trends, all of the results hold—as reported in Table 5. The point estimates are even larger. For the

Depression Index, the coefficient in column (1) suggests that one more degree of interstate bank

branch openness reduces the Depression Index by 0.522, a 15% drop of the sample mean. When

examining severe depression among the low income group (D-Depression Index Above 8), the

estimated coefficient in column (2) indicates that one more degree of interstate branch

deregulation is associated with a drop in the probability of severe mental depression of about

3.8%. In the remainder of the analyses, we include state-specific linear time trends but all of the

results hold if these trends are excluded.

4.3 Spatial Regression Discontinuity Design

18

We next employ a spatial regression discontinuity design to address concerns that omitted

variables bias the results. As argued by Huang (2008), geographically proximate counties across

a shared state border are likely to be similar in both observable variables (e.g., economic growth)

and unobservable—and hence omitted—variables. As a result, adjacent counties with different

bank branch regulations can be a good “match” with respect to observables and unobservables

and therefore provide an additional empirical strategy for assessing the impact of changes in

credit conditions on mental health. Employing this strategy, we compare changes in mental

health in adjacent counties across state borders with different and changing branching regulations.

To implement the empirical design, we obtain 1990 state- and county-boundary data from

the U.S. Census and create contiguous county-pairs across state borders.5 We then map this

county-pair data into our NLSY79 sample and re-perform the main tests above. We restrict the

sample to contiguous county-pairs in these analyses. We control for county-pair and gender-race-

year fixed effects, as well as state-specific linear time trends, in all regressions and cluster the

standard errors at the county-pair level.

As shown in Table 6, the results from spatial regression discontinuity analyses confirm

the analyses above: interstate branch deregulation is associated with a sharp reduction in mental

illness among low-income individuals, but not among high-income individuals. Among the low-

income group distributed in contiguous counties along the state border, an increase in bank

branch deregulation is associated with a reduction in mental depression.6

5. Evidence on the Channels between Bank Branch Deregulation and Depression 5 https://www.census.gov/geo/maps-data/data/tiger-cart-boundary.html 6 One potential concern with these analyses is that because cross-state lending is not restricted, residents close to the borders with other states may turn to those banking markets for credit. In this case, the bank regulations of the state in which an individual is a resident might not matter much. Since people tend to borrow from geographically close banks, we examine whether the reported results hold when eliminating county-pairs across state borders that are “too close.” To define “too close,” we first note that Petersen and Rajan (2002) find (in their Table I, Panel A) that the median distance between a firm and its lender is nine miles. Starting with this definition of “too close,” we first measure the travel distance between the geographic centers of two contiguous counties and then drop county-pairs in which the distance is less than nine miles. We try different cutoffs (see Table OA3) and confirm that the results presented in Table 6 hold across different cutoff levels of closeness.

19

In this section, we use additional empirical strategies to provide evidence on the channels

connecting interstate bank branch deregulation and mental depression. We first consider the

“firm-credit channel,” which posits that interstate bank branch deregulation eased firm credit

constraints, which in turn boosted income and employment among workers, with positive

ramifications on mental health. As demonstrated by Rice and Strahan (2010), interstate bank

branch deregulation eased firm credit constraints. As shown above, interstate bank branch

deregulation reduced mental depression among low-income workers. We now assess whether

income, employment, and mental health improved primarily among individuals most likely to be

affected by an easing of firm credit triggered by interstate bank branch deregulation.

To explore the firm-credit channel, we differentiate counties by the degree to which firms

in each county depend on external finance, where banks are the major providers of external

finance. The motivation is as follows: If branch deregulation improves mental health by easing

firm credit constraints and enhancing the economic conditions of workers, then we should find

an especially pronounced impact of deregulation on the economic conditions and mental health

of workers in counties where bank branch deregulation is likely to have the biggest effect on

firms: counties with firms that depend heavily on banks for external finance. To examine this

channel, we follow Rajan and Zingales (1998) and differentiate counties by the degree to which

firms are financially dependent. We then examine whether interstate branch deregulation exerts

an especially pronounced effect on mental health, employment, and income among low-income

workers in counties dominated by financially-dependent firms. As a robustness check, we

differentiate counties by the density of bank branches before bank branch deregulation. Holding

other features of the counties constant, we expect that bank branch deregulation will have a

bigger effect on economic conditions and mental health in “under-banked” counties than in

counties that are already densely banked before deregulation.

Second, we consider the “consumer-credit channel.” The consumer-credit channel

suggests that bank deregulation eases individual credit constraints, which in turn allows

individuals to reduce economic stress, with positive ramifications on mental health. To shed

20

some empirical light on the consumer-credit channel, we assess whether interstate bank branch

deregulation triggers increased borrowing among the same group of individuals who experience

a drop in mental depression: low-income workers.

5.1 Firm-credit Channel: External Financial Dependence (EFD)

We now examine whether interstate bank branch deregulation triggered improvements in

income, employment, and mental health among individuals in counties where bank branch

deregulation is most likely to ease firm-credit constraints: counties with firms that depend

heavily on banks for external finance. We divide counties into those with firms that depend

heavily on external finance dependence (EFD) and those that do not. To obtain county-level EFD,

we first follow prior studies (Rajan and Zingales, 1998; Duchin, Ozbas and Sensoy, 2010) and

define each firm’s EFD as the ratio of capital expenditures minus net cash flows from operating

activities over the capital expenditures. We obtain data from Compustat and calculate the firm-

year EFD over a long time-series window 1980-2000. We first aggregate firm-level values to the

3-digit SIC industry-level and calculate annual EFD for that industry. We next take the median

of industry-year EFD over the long window (1980-2000) as the measurement of industry-level

EFD. To arrive at the county-level EFD, we retrieve data from County Business Patterns (CBP)

administered by the U.S. Census and compute the employment-weighted EFD across all 3-digit

SIC industries for each county-year. Finally, we average the county-year EFD over the pre-shock

period 1989-1991 to gauge a persistent measurement of county-level EFD. We split the low

income group sample into low county EFD and high county EFD based on the county EFD

median. The results are reported in Table 7.

In line with the firm-credit channel view, we find that low-income individuals living in

high EFD counties experience large improvements in mental health, employment, and income

following interstate branch deregulation but low-income individuals living in low EFD counties

do not. When focusing on low-income individuals in counties most likely to be influenced by

21

deregulation (high EFD counties), the estimated effect of branch deregulation on mental health is

large. From column (1) of Table 7, one more degree of interstate bank branch deregulation is

associated with a reduction in the Depression Index of 0.826 among low-income individuals in

high EFD counties, which represents a 23.5% drop relative to the sample mean. Table 7 results

are consistent with the view that interstate branch deregulation eases credit conditions among

firms that depend on banks for external finance and that this easing of credit constraints

improves the economic conditions of workers with positive repercussion on mental health.

Furthermore, consistent with the findings in Beck, Levine, and Levkov (2010) that

improvements in bank operations exert a disproportionately positive influence on low-income

workers, we find that the positive effect of interstate branch deregulation on employment,

income, and mental health occurs among low-income individuals in high EFD counties.

We also differentiate counties by bank density. The hypothesized economic mechanism is

as follows: The impact of interstate branch deregulation on firm-credit conditions in a county

will be larger if the county had lower bank density prior to interstate branch deregulation than if

the county was already heavily banked. Namely, since deregulation will tend to have bigger

positive effect on the availability of banking services and credit in previously under-banked

counties than in already densely-banked counties, deregulation will improve employment,

income, and mental health more in counties with low-density banking prior to deregulation. To

assess this view, we compile county-level branch density data, which is available from June 1994,

from the Sum of Deposits (SOD) files supported by Federal Deposit Insurance Corporate (FDIC).

We use pre-deregulation branch data from 1994 and compute the total number of branches per

county scaled by total county population. We split counties into low- and high-density counties

based on the median of this branch per capita indicator.

These analyses also yield results consistent with the firm-credit channel view. As shown

in Table 8, we find that low-income individuals living in low pre-deregulation branch density

counties experience large improvements in employment, income, and mental health following

interstate branch deregulation. In contrast, low-income individuals living in high pre-

22

deregulation branch density counties do not experience significant gains in employment, income

or mental health. As with the county-level EFD results, these results on pre-deregulation branch

density at the county-level also support the firm-credit channel.

5.2 Consumer-Credit Channel

We now consider the “consumer-credit channel.” Applied to our study of branch

deregulation and mental health, this channel suggests that interstate bank branch deregulation

improves credit conditions for individuals, so that they can borrow to smooth the vagaries of

economic life, reducing stress and mental illness. We shed empirical light on this potential

channel from interstate branch deregulation to mental depression by using two measures of

individual borrowing. Leverage equals the ratio of total liabilities excluding mortgages on

residential property (e.g., debts on farm/business/other property, debts on vehicles etc.) over total

assets excluding the market value of residential property (e.g., total amount of money assets like

savings, market value of farm/business/other property, market value of all vehicles and all other

assets worth more than $500 etc.). D/I Ratio is the ratio of total liabilities excluding mortgages

on residential property as defined above over respondent’s total income (including wages,

salaries, and income from farm or business).

As shown in Table 9, we find no evidence that interstate branch deregulation facilitates

borrowing by individuals most affected by deregulation: low-income workers. We also show that

the lack of an association between interstate branch deregulation and borrowing by individuals

holds in low- and high-EFD counties and in low and high pre-deregulation branch density

counties.

23

6. Conclusion

In this paper, we examine the question: Do bank regulatory reforms that improve local

credit conditions reduce the incidence of mental depression? Research suggests both that (1) the

removal of restrictions on interstate bank branching eases credit conditions, spurs

entrepreneurship, and boosts earnings, especially the earnings of low-income workers, and (2)

economic duress, including low-incomes and unemployment, tends to increase mental depression.

Combined, these strands of research motivate the hypothesis that we examine: Bank regulatory

reforms that improve economic conditions will tend to reduce mental depression, especially

among low-income workers. Since improvement in the operation of banking systems tends to

spur “churn,” the more rapid entry and exit of firms, this can create uncertainty and anxiety

among workers that increase mental distress.

We discover that the lowering of regulatory impediments to interstate branching reduced

mental health disease among low-income households, especially in counties dominated by firms

that depend heavily on banks. These results also hold when using a spatial regression

discontinuity design to compare contiguous counties across state borders with different interstate

branch regulations. On the mechanisms linking credit conditions and mental health, we find that

interstate branch deregulation boosted employment and income among low-income workers in

counties dominated by bank-dependent firms, which is consistent with the firm-credit channel

argument that interstate bank branch deregulation eased firm credit constraints, which in turn

boosted income and employment among workers, with positive ramifications on mental health.

24

References

Adhvaryu, Achyuta, Fenske, James, and Anant Nyshadham, 2019, Early life circumstance and

adult mental health, Journal of Political Economy, forthcoming.

Agarwal, Sumit, Souphala Chomsisengphet, Neale Mahoney, and Johannes Stroebel, 2018, Do

banks pass through credit expansions to consumers who want to borrow?, The Quarterly

Journal of Economics 133, 129-190.

Agarwal, Sumit, and Wenlan Qian, 2014, Consumption and debt response to unanticipated

income shocks: evidence from a natural Experiment in Singapore, American Economic

Review 104, 4205-4230.

Altonji, Joseph G, and Charles R Pierret, 2001, Employer learning and statistical discrimination,

The Quarterly Journal of Economics 116, 313–350.

Apouey, Bénédicte, and Andrew E Clark, 2015, Winning big but feeling no better? the effect of

lottery prizes on physical and mental health, Health Economics 24, 516–538.

Bartel, Ann, and Paul Taubman, 1986, Some economic and demographic consequences of

mental illness. Journal of Labor Economics, 4(2), pp.243-256.

Beck, Thorsten, Ross Levine, and Alexey Levkov, 2010, Big bad banks? the winners and losers

from bank deregulation in the united states, The Journal of Finance 65, 1637–1667.

Berger, Lawrence M., J. Michael Collins, and Laura Cuesta, 2016, Household debt and adult

depressive symptoms in the United States, Journal of Family and Economic Issues 37,

42-57.

Black, Sandra E, and Philip E Strahan, 2001, The division of spoils: rent-sharing and discrimination in a regulated industry. American Economic Review 91, 814-831.

Black, Sandra E, and Philip E Strahan, 2002, Entrepreneurship and bank credit availability, The

Journal of Finance 57, 2807–2833.

Chetty, Raj, Nathaniel Hendren, Patrick Kline, and Emmanuel Saez, 2014, Where is the land of

opportunity? the geography of intergenerational mobility in the united states, The

Quarterly Journal of Economics 129, 1553–1623.

Chisholm, D., Sweeny, K., Sheehan, P., Rasmussen, B., Smit, F., Cuijpers, P. and Saxena, S.,

2016. Scaling-up treatment of depression and anxiety: a global return on investment

analysis. The Lancet Psychiatry, 3(5), pp.415-424.

25

Clayton, Maya, Jose Linares-Zegarra, and John O.S. Wilson, 2015, Does debt affect health?

Cross-country evidence on the debt-health nexus. Social Science & Medicine 130, 51-58.

Dolan, Paul, Tessa Peasgood, and Mathew White, 2008, Do we really know what makes us

happy? a review of the economic literature on the factors associated with subjective well-

being, Journal of Economic Psychology 29, 94–122.

Duchin, Ran, Oguzhan Ozbas, and Berk A. Sensoy, 2010, Costly external finance, corporate

investment, and the subprime mortgage credit crisis, Journal of Financial Economics 97,

418–435.

Favara, Giovanni, and Jean Imbs, 2015, Credit supply and the price of housing, American

Economic Review 105, 958–92.

Garmaise, Mark J., and Tobias J. Moskowitz, 2006 Bank mergers and crime: The real and social

effects of credit market competition. The Journal of Finance 61, 495-538.

Gili, Margalida, Miquel Roca, Sanjay Basu, Martin McKee, and David Stuckler, 2012, The

mental health risks of economic crisis in Spain: evidence from primary care centres, 2006

and 2010. The European Journal of Public Health 23, 103-108.

Huang, Rocco R, 2008, Evaluating the real effect of bank branching deregulation: Comparing

contiguous counties across us state borders, Journal of Financial Economics 87, 678–705.

Insel, Thomas R., 2008, Assessing the economic costs of serious mental illness, American

Journal of Psychiatry 165, 663–665.

Jacoby, Hanan G., 1994, Borrowing constraints and progress through school: Evidence from

Peru, Review of Economics and Statistics 76, 151-160.

Jayaratne, Jith, and Philip E Strahan, 1996, The finance-growth nexus: Evidence from bank

branch deregulation, The Quarterly Journal of Economics 111, 639–670.

Johnson, Christian A., and Tara Rice, 2008, Assessing a decode of interstate bank branching,

Washington and Lee Law Review 65, 73-127.

Kerr, William R, and Ramana Nanda, 2009, Democratizing entry: Banking deregulations,

financing constraints, and entrepreneurship, Journal of Financial Economics 94, 124–149.

Kessler, Ronald C., 2012, The costs of depression. Psychiatric Clinics, 35(1), pp.1-14.

King, Robert G., and Ross Levine, 1993, Finance and growth: Schumpeter might be right, The

Quarterly Journal of Economics 108, 717-737.

26

Levine, Ross, and Sara Zervos, 1998, Stock markets, banks, and economic growth, American

Economic Review 88, 537-558.

Levine, Ross, Alex Levkov, and Yona Rubinstein, 2014, Bank deregulation and racial inequality

in America. Critical Finance Review 3, 1-48.

Levine, Ross, and Yona Rubinstein, 2017, Smart and illicit: Who becomes an entrepreneur and

do they earn more?, The Quarterly Journal of Economics 132, 963–1018.

Levine, Ross, and Yona Rubinstein, 2018, Selection into entrepreneurship and self-employment,

National Bureau of Economic Research Working Paper No. 25350.

Levine, Stephen Z, 2013, Evaluating the seven-item center for epidemiologic studies depression

scale short-form: A longitudinal us community study, Social Psychiatry and Psychiatric

Epidemiology 48, 1519–1526.

Lund, Crick, Alison Breen, Alan J Flisher, Ritsuko Kakuma, Joanne Corrigall, John A Joska,

Leslie Swartz, and Vikram Patel, 2010, Poverty and common mental disorders in low and

middle income countries: A systematic review, Social Science & Medicine 71, 517–528.

Lund, Crick, Mary De Silva, Sophie Plagerson, Sara Cooper, Dan Chisholm, Jishnu Das, Martin

Knapp, and Vikram Patel, 2011, Poverty and mental disorders: Breaking the cycle in low-

income and middle-income countries, The Lancet 378, 1502–1514.

Miller, Douglas L., Marianne E. Page, Ann Huff Stevens, and Mateusz Filipski, 2009, Why are

recessions good for your health? American Economic Review 99, 122-27.

Pearlin, Leonard I., Elizabeth G. Menaghan, Morton A. Lieberman, and Joseph T. Mullan, 1981,

The stress process, Journal of Health and Social Behavior 22, 337-356.

Persson, Petra, and Maya Rossin-Slater, 2018, Family ruptures, stress, and the mental health of

the next generation, American Economic Review 108, 1214–52.

Petersen, Mitchell A, and Raghuram G Rajan, 2002, Does distance still matter? the information

revolution in small business lending, The Journal of Finance 57, 2533–2570.

Popov, A. 2014, Credit constraints and investment in human capital: Training evidence from

transition economies. Journal of Financial Intermediation, 23, 76.100.

Popov, A., 2018. Evidence on finance and economic growth. In Thorsten Beck and Ross Levine,

editors, Handbook of Finance and Development, Edward Elgar – Cheltenham, United

Kingdom.

27

Rajan, Raghuram G., and Luigi Zingales, 1998, Financial dependence and growth, American

Economic Review 88, 559–586.

Ramchandani, Paul, Alan Stein, Jonathan Evans, Thomas G. O'Connor, and ALSPAC Study

Team, 2005, Paternal depression in the postnatal period and child development: a

prospective population study. The Lancet 365, 2201-2205.

Reeves, Aaron, David Stuckler, Martin McKee, David Gunnell, Shu-Sen Chang, and Sanjay

Basu, 2012, Increase in state suicide rates in the USA during economic recession, The

Lancet 380, 1813–1814.

Rice, Tara, and Philip E Strahan, 2010, Does credit competition affect small-firm finance?, The

Journal of Finance 65, 861–889.

Ruhm, Christopher J., 2000, Are recessions good for your health? The Quarterly Journal of

Economics 115, 617-650.

Ruhm, Christopher J., 2005, Healthy living in hard times. Journal of Health Economics 24, 341-

363.

Ruhm, Christopher J., 2015, Recessions, healthy no more? Journal of Health Economics 42, 17-

28.

Sun, Stephen Teng, and Constantine Yannelis, 2016, Constraints, credit and demand for higher

education: Evidence from financial deregulation, Review of Economics and Statistics 98,

12–24.

Wickham, S., Whitehead, M., Taylor-Robinson, D. and Barr, B., 2017. The effect of a transition

into poverty on child and maternal mental health: a longitudinal analysis of the UK

Millennium Cohort Study. The Lancet Public Health, 2(3), pp.e141-e148.

World Health Organization, 2011, Global status report on non-communicable diseases

2010. Geneva: WHO.

World Health Organization, 2015, Mental Health Atlas 2014, Geneva: WHO.

World Health Organization, 2018, Mental Health Atlas 2017, Geneva: WHO.

28

Figure 1. Annual Trend of CES-Depression Index

This figure plots the time-series trend of the total score of CES-Depression Index over the period 1992 to 2014. Total score ranges from 0 (least depressed) to 21 (most depressed), measuring the severity of individuals’ mental health distress. The data sources from NLSY79 panels surveys administered by the BLS.

23

45

1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014Year

Depression Index

29

Figure 2. Heterogeneous Evolution of CES-Depression Index by Degree of Financial Deregulation

This figure compares heterogeneous evolution of CES-Depression Index within low-income individuals residing in more or less bank branching deregulated states. The NLSY79 surveys each individual’s mental health in 1992, 1994, and the year immediately after the persons becomes 40 years old (over 1998-2005). Thus, we have two depression measures before the 1994 Riegle-Neal Act and one measure afterwards.

30

Table 1. Summary Statistics

This table summarizes descriptive statistics for the main variables. The sample period ranges from 1992 to 2005, in line with the data availability for our key mental health variables. Each respondent is surveyed three times over the depression symptoms, the first in 1992, the second in 1994 and the final when (s)he turns age 40 (over 1998-2005). Our variable of interest, IBBEA Index, takes the value from 0 (most restrictive) to 4 (most deregulated). All other variables are defined in Appendix 1.

Variable N Mean S.D. Min Median Max State-Level Bank Branching Deregulation IBBEA Index 714 1.307 1.479 0 1 4 CES-Depression Scale Depression Index 14,299 3.508 3.960 0.000 2.000 21.000 D-Depression Index Above 8 14,299 0.136 0.342 0.000 0.000 1.000 D-Depression Index Above 10 14,299 0.082 0.275 0.000 0.000 1.000 D-Depression Index Above 12 14,299 0.052 0.223 0.000 0.000 1.000 % of Answers “0” 14,299 0.676 0.298 0.000 0.714 1.000 % of Answers “1” 14,299 0.200 0.216 0.000 0.143 1.000 % of Answers “2” 14,299 0.072 0.138 0.000 0.000 1.000 % of Answers “3” 14,299 0.053 0.138 0.000 0.000 1.000 Trouble Eating 14,299 0.332 0.732 0.000 0.000 3.000 Trouble Concentrating 14,299 0.527 0.822 0.000 0.000 3.000 Feel Depressed 14,299 0.385 0.747 0.000 0.000 3.000 Everything an Effort 14,299 0.658 0.988 0.000 0.000 3.000 Trouble sleeping 14,299 0.728 0.983 0.000 0.000 3.000 Feel Sad 14,299 0.386 0.720 0.000 0.000 3.000 Lethargic 14,299 0.493 0.779 0.000 0.000 3.000 Labor Market Outcomes D-WorkHours 16,107 0.790 0.408 0.000 1.000 1.000 Log(WorkHours) 16,107 5.778 3.051 0.000 7.418 8.269 Log(EmpWeeks) 16,264 2.991 1.588 0.000 3.970 3.970 Log(TotalIncome) 15,443 7.574 4.183 0.000 9.699 11.963 Leverage Ratios Leverage 12,154 0.671 1.751 0.000 0.146 12.000 D/I Ratio 10,361 0.695 1.558 0.000 0.151 8.338 State-Level Controls Log(GSP Per Capita) 714 10.64 0.248 10.168 10.618 12.041 Log(Population) 714 15.008 1.031 13.052 15.139 17.403

31

Table 2. Bank Branching Deregulation and Total Score of CES-Depression Scale

This table reports OLS regression results of the effect of the interstate bank branching deregulation on total score of CES- Depression Scale over 1992 to 2005. Each respondent has three observations for Depression Index, one in 1992, one in 1994 and one when (s)he turns age 40. Column (1) to (3) report the results conditioning on state and gender-race-year fixed effects, and column (4) to (6) further include individual fixed effects. Column (1) and (4) report the results in the full sample, and column (2) and (3), (5) and (6) report results in the low- and high-income group respectively. The dependent variable is Depression Index, ranging from 0 (least depressed) to 21 (most depressed). The key independent variable is IBBEA Index, ranging from 0 (most restrictive) to 4 (most deregulated). Education denotes the highest grade/years of schooling completed by the respondent, ranging from 0 (never attend school) to 20 (8th year college or more). Age is the difference between the birth year and the survey year. FamilyIncome7981 is the natural logarithm of one plus the average total family income (in 2010 dollars) over 1979 to 1981. MotherEducation and FatherEducation measure maximum years of schooling received by the respondent's parents. D-LiveWithParents equals one if the respondent lived with both parents at the age 14. AFQTAbilityTest is the normalized percentile score after controlling for age groups among all the respondents in the 1980 survey. A higher percentile indicates better learning aptitude. RosenbergSelfEsteem is an aggregate score ranging from 0 to 30 in the 1987 survey to measure the respondent's self-evaluation of personal worth. A higher value indicates a higher level of self-esteem. PearlinSenseOfControl is an aggregate score ranging from 7 to 28 in the 1992 survey to measure the respondent's sense of control over one's life. A higher value indicates a more positive self-concept and a stronger internal control over one's destiny. D-RiskAttitude is a dummy variable that equals one if the respondent answers “Yes” facing the question “Would you take job that could either double family income or cut income by a third” in the 1993 survey, and zero otherwise. State controls include log of GSP per capita and log of total population. All regressions use the sample weights provided by NLSY79. Heteroskedasticity-robust standard errors clustered at the state level are in parentheses. We test the coefficient difference between the high- and low-income group and p-value is reported in the last row. *, **, and *** indicate significance at 10%, 5%, and 1%.

32

Full Sample Low Income High Income Full Sample Low Income High Income

Depression Index

Depression Index

Depression Index

Depression Index

Depression Index

Depression Index

(1) (2) (3) (4) (5) (6) IBBEA Index -0.140** -0.284** -0.0531 -0.119** -0.297** 0.00689

(-2.34) (-3.19) (-0.86) (-2.29) (-3.21) (0.14)

Education -0.0637** -0.0179 -0.0824**

(-2.74) (-0.61) (-2.88)

Age -0.0259 -0.00650 -0.0263

(-1.31) (-0.16) (-1.21)

FamilyIncome7981 -0.253*** -0.335** -0.109

(-3.55) (-2.76) (-1.42)

MotherEducation -0.00129 0.0130 -0.00830

(-0.06) (0.38) (-0.35)

FatherEducation 0.0165 0.0230 0.00629

(0.83) (0.62) (0.35)

D-BothParents -0.383** -0.580** -0.211

(-2.56) (-2.57) (-1.31)

AFQTAbilityTest -1.116*** -1.439*** -0.553**

(-5.59) (-5.00) (-2.12)

RosenbergSelfEsteem -0.0898*** -0.109*** -0.0562**

(-7.29) (-5.42) (-3.11)

PearlinSenseOfControl -0.243*** -0.302*** -0.192***

(-12.15) (-8.59) (-9.29)

D-RiskAttitude 0.277** 0.361** 0.202*

(2.71) (2.43) (1.69)

Log(GSP Per Capita) 0.661 0.783 0.252 -0.452 -1.710 0.745

(0.42) (0.32) (0.20) (-0.33) (-0.77) (0.61)

Log(Population) -0.458 -0.786 -0.208 -0.257 -0.676 0.0706 (-1.42) (-1.43) (-0.57) (-0.68) (-0.83) (0.25) State FE Yes Yes Yes

33

Individual FE Yes Yes Yes Gender-Race-Year FE Yes Yes Yes Yes Yes Yes Observation 14299 6612 7687 14299 6612 7687 R-Squared 0.133 0.123 0.0997 0.377 0.375 0.326 P-Value 0.016 0.004

34

Table 3. Bank Branching Deregulation and Answers’ Distribution of CES-Depression Scale

This table reports OLS regression results of the effect of the interstate bank branching deregulation on the choice distribution of the CES-Depression Scale in the low income group over 1992 to 2005. D-Depression Index Above 8, D-Depression Index Above 10, D-Depression Index Above 12 respectively represents a dummy variable that equals one if Depression Index is above the cutoff 8/10/12, and zero otherwise. This set of dummies can measure the likelihood of clinical major depression. CES-Depression Scale is made up of seven statements evaluating the level of depression symptoms over the last week and respondents choose from the following four alternatives as their ratings: (1) Rarely/None of the time/1 Day (respondents choose 0 as the answer), (2) Some/A little of the time/1-2 Days (respondents choose 1 as the answer), (3) Occasionally/Moderate amount of the time/3-4 Days (respondents choose 2 as the answer), (4) Most/All of the times/5-7 Days (respondents choose 3 as the answer). % of Answers “0,” % of Answers “1,” % of Answers “2” and % of Answers “3” respectively denotes the percentage of choosing 0/1/2/3 as the ratings among the seven statements. The key independent variable is IBBEA Index, ranging from 0 (most restrictive) to 4 (most deregulated). State controls include log of GSP per capita and log of total population. We include individual and gender-race-year fixed effects in all columns. All regressions use the sample weights provided by NLSY79. Heteroskedasticity-robust standard errors clustered at the state level are in parentheses. *, **, and *** indicate significance at 10%, 5%, and 1%.

Low Income Group

D-Depression Index Above 8

D-Depression Index Above 10

D-Depression Index Above 12

% of Answers “0”

% of Answers “1”

% of Answers “2”

% of Answers “3”

(1) (2) (3) (4) (5) (6) (7) IBBEA Index -0.0145* -0.0178** -0.0142** 0.0167** -0.00177 -0.00412 -0.0108**

(-1.82) (-2.54) (-2.40) (3.40) (-0.37) (-1.46) (-2.58)

Log(GSP Per Capita) -0.113 -0.197 -0.111 0.0406 0.114 -0.106 -0.0487

(-0.54) (-0.94) (-0.54) (0.25) (0.80) (-1.41) (-0.58)

Log(Population) -0.0579 -0.0536 -0.0399 0.0448 -0.0156 -0.00685 -0.0224 (-0.97) (-1.27) (-0.89) (1.04) (-0.66) (-0.43) (-0.63) Individual FE Yes Yes Yes Yes Yes Yes Yes Gender-Race-Year FE Yes Yes Yes Yes Yes Yes Yes Observation 6612 6612 6612 6612 6612 6612 6612 R-Squared 0.255 0.225 0.179 0.401 0.224 0.146 0.259

35

Table 4. Bank Branching Deregulation and Seven Dimensions of CES-Depression Scale

This table reports OLS regression results of the effect of the interstate bank branching deregulation on seven dimensions of the CES-Depression Scale in the low income group over 1992 to 2005. CES-Depression Scale is made up of seven statements evaluating the level of depression symptoms over the last week and respondents choose from the following four alternatives as their ratings: (1) Rarely/None of the time/1 Day (respondents choose 0 as the answer), (2) Some/A little of the time/1-2 Days (respondents choose 1 as the answer), (3) Occasionally/Moderate amount of the time/3-4 Days (respondents choose 2 as the answer), (4) Most/All of the times/5-7 Days (respondents choose 3 as the answer). Trouble Eating is “I did not feel like eating; my appetite was poor”. Trouble Concentrating is “I had trouble keeping my mind on what I was doing”. Feel Depressed is “I feel depressed”. Everything an Effort is “I felt that everything I did was an effort”. Trouble sleeping is “My sleep was restless”. Feel Sad is “I felt sad”. Lethargic is “I could not get ‘going’”. The key independent variable is IBBEA Index, ranging from 0 (most restrictive) to 4 (most deregulated). State controls include log of GSP per capita and log of total population. We include individual and gender-race-year fixed effects in all columns. All regressions use the sample weights provided by NLSY79. Heteroskedasticity-robust standard errors clustered at the state level are in parentheses. *, **, and *** indicate significance at 10%, 5%, and 1%.

Low Income Group

Trouble Eating

Trouble Concentrating

Feel Depressed

Everything an Effort

Trouble sleeping Feel Sad Lethargic

(1) (2) (3) (4) (5) (6) (7) IBBEA Index -0.0131 -0.0771*** -0.0566** -0.0348** -0.0509** -0.0498** -0.0151

(-0.97) (-4.15) (-3.16) (-2.12) (-2.10) (-2.85) (-0.73)

Log(GSP Per Capita) 0.0548 -0.493 -0.0490 0.272 -0.372 -0.431 -0.693

(0.18) (-1.15) (-0.10) (0.65) (-0.51) (-1.06) (-1.66)

Log(Population) 0.00874 -0.168* -0.0611 -0.323** -0.114 -0.103 0.0848 (0.05) (-1.76) (-0.45) (-2.63) (-0.58) (-1.03) (0.74) Individual FE Yes Yes Yes Yes Yes Yes Yes Gender-Race-Year FE Yes Yes Yes Yes Yes Yes Yes Observation 6612 6612 6612 6612 6612 6612 6612 R-Squared 0.268 0.230 0.301 0.279 0.255 0.240 0.239

36

Table 5. Bank Branching Deregulation and CES-Depression Scale-Robustness Checks

This table reports robustness checks on OLS regression results of the effect of the interstate bank branching deregulation on the CES-Depression Scale in the low income group over 1992 to 2005, by further considering state-level linear time trends. Depression Index ranges from 0 (least depressed) to 21 (most depressed). D-Depression Index Above 8, D-Depression Index Above 10, D-Depression Index Above 12 respectively represents a dummy variable that equals one if Depression Index is above the cutoff 8/10/12, and zero otherwise. This set of dummies can measure the likelihood of clinical major depression. The key independent variable is IBBEA Index, ranging from 0 (most restrictive) to 4 (most deregulated). State controls include log of GSP per capita and log of total population. We include individual and gender-race-year fixed effects, as well as state-level linear time trends in all columns. All regressions use the sample weights provided by NLSY79. Heteroskedasticity-robust standard errors clustered at the state level are in parentheses. *, **, and *** indicate significance at 10%, 5%, and 1%.

Low Income Group

Depression Index

D-Depression Index Above 8

D-Depression Index Above 10

D-Depression Index Above 12

(1) (2) (3) (4) IBBEA Index -0.522** -0.0380** -0.0309* -0.0253*

(-3.19) (-2.07) (-1.73) (-1.71)

Log(GSP Per Capita) 1.107 -0.0242 -0.0411 0.109

(0.41) (-0.09) (-0.18) (0.46)

Log(Population) -1.341* -0.0813 -0.0841* -0.0752 (-1.70) (-1.29) (-1.96) (-1.64) Individual FE Yes Yes Yes Yes Gender-Race-Year FE Yes Yes Yes Yes StateLinearTrends Yes Yes Yes Yes Observation 6612 6612 6612 6612 R-Squared 0.381 0.259 0.228 0.184

37

Table 6. Spatial Regression Discontinuity Design

This table presents OLS regression results of the effect of the interstate bank branching deregulation on the CES-Depression Scale over 1992 to 2005. Odd columns report the results in the low income group and even columns report the results in the high income. We exploit a spatial regression discontinuity design and restrict the sample to the contiguous county-pairs across the state borders. Depression Index ranges from 0 (least depressed) to 21 (most depressed) and measures the severity of depression symptoms. D-Depression Index Above 8, D-Depression Index Above 10, D-Depression Index Above 12 respectively represents a dummy variable that equals one if Depression Index is above the cutoff 8/10/12, and zero otherwise. This set of dummies can measure the likelihood of clinical major depression. The key independent variable is IBBEA Index, ranging from 0 (most restrictive) to 4 (most deregulated). State controls include log of GSP per capita and log of total population. We include county-pair and gender-race-year fixed effects, as well as state-level linear time trends in all columns. Heteroskedasticity-robust standard errors clustered at the county-pair level are in parentheses. We test the coefficient difference between the high- and low-income and p-value is reported in the last row. *, **, and *** indicate significance at 10%, 5%, and 1%.

Depression Index D-Depression Index Above 8

D-Depression Index Above 10

D-Depression Index Above 12

Low Income

High Income

Low Income

High Income

Low Income

High Income

Low Income

High Income

(1) (2) (3) (4) (5) (6) (7) (8) IBBEA Index -0.300** 0.0757 -0.0284*** -0.000380 -0.0297*** 0.00665 -0.0272*** 0.00231

(-2.86) (0.92) (-3.37) (-0.05) (-3.71) (1.07) (-3.80) (0.48)

Log(GSP Per Capita) 0.350 -0.341 0.0179 0.243 0.00437 0.263** -0.0709 0.0823

(0.27) (-0.21) (0.14) (1.61) (0.05) (2.58) (-1.07) (1.04)

Log(Population) -0.988*** 0.699** -0.0742*** -0.0193 -0.0607** -0.00839 -0.0730** -0.000907 (-3.95) (2.16) (-3.67) (-0.93) (-2.43) (-0.41) (-2.78) (-0.05) County-Pair FE Yes Yes Yes Yes Yes Yes Yes Yes Gender-Race-Year FE Yes Yes Yes Yes Yes Yes Yes Yes StateLinearTrends Yes Yes Yes Yes Yes Yes Yes Yes Observation 4521 4889 4521 4889 4521 4889 4521 4889 R-Squared 0.219 0.131 0.172 0.0843 0.156 0.0558 0.122 0.0516 P-Value 0.003 0.007 0.001 0.001

38

Table 7. Labor Market Outcomes: Cross-County Heterogeneity by Local External Finance Dependence

This table presents OLS regression results of the firm-credit channel tests by exploiting the heterogeneous responses in the low income group between residents living in counties with high` and low external finance dependence (EFD). Following the literature (Rajan and Zingales 1998; Duchin, Ozbas and Sensoy 2010), we define EFD as the ratio of capital expenditures minus net cash flows from operating activities over the capital expenditures. We first aggregate firm-level values to the 3-digit SIC industry-level and calculate annual EFD for each industry. We next take the median of industry-year EFD over the long window (1980-2000) as the measurement of industry-level EFD. To arrive at the county-level EFD, we compute the employment-weighted EFD across all 3-digit SIC industries for each county-year. Then we average the county-year EFD over the pre-shock period 1989-1991 to gauge a persistent measurement of county-level EFD. We split the low income group sample into low county EFD and high county EFD based on the county EFD median. Depression Index ranges from 0 (least depressed) to 21 (most depressed) and measures the severity of depression symptoms. D-WorkHours is a dummy variable that equals one if the respondent participates in the job market and has positive working hours in a given year, and zero otherwise. Log(WorkHours) denotes natural logarithm of one plus total number of hours worked in a given year. Log(EmptWeeks) denotes the natural logarithm of one plus total employment weeks in a given year. Log(TotalIncome) is the natural logarithm of one plus the respondent’s total income (in 2010 dollars), including wages, salaries, and income from farm or business. The key independent variable is IBBEA Index, ranging from 0 (most restrictive) to 4 (most deregulated). State controls include log of GSP per capita and log of total population. We include individual and gender-race-year fixed effects, as well as state-level linear time trends in all columns. All regressions use the sample weights provided by NLSY79. Heteroskedasticity-robust standard errors clustered at the state level are in parentheses. We test the coefficient difference between the high and low EFD counties and p-value is reported in the last row. *, **, and *** indicate significance at 10%, 5%, and 1%.

39

Depression Index D-WorkHours Log(WorkHours) Log(EmpWeeks) Log(TotalIncome)

High EFD

Low EFD

High EFD

Low EFD

High EFD

Low EFD

High EFD

Low EFD

High EFD

Low EFD

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) IBBEA Index -0.826** -0.225 0.0193** 0.00148 0.130** 0.0153 0.0730** -0.000697 0.205** -0.00390

(-2.56) (-1.14) (2.71) (0.23) (2.35) (0.37) (2.38) (-0.03) (2.42) (-0.06)

Log(GSP Per Capita) 1.019 -0.0186 0.251 0.142 2.284 1.529 1.178 0.907 2.051 2.263

(0.32) (-0.00) (1.26) (0.66) (1.59) (1.18) (1.56) (1.35) (0.93) (1.13)

Log(Population) -1.378 -0.889 0.0266 -0.0974 -0.121 -0.193 -0.121 -0.0205 -0.192 -4.501 (-1.07) (-1.10) (0.71) (-0.93) (-0.40) (-0.30) (-0.76) (-0.07) (-0.47) (-1.45) Individual FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Gender-Race-Year FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes StateLinearTrends Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Observation 3098 3479 7537 8503 7537 8503 7610 8587 7220 8164 R-Squared 0.410 0.356 0.490 0.492 0.532 0.533 0.521 0.521 0.535 0.553 P-Value 0.147 0.017 0.025 0.010 0.024

40

Table 8. Labor Market Outcomes: Cross-County Heterogeneity by Local Credit Condition

This table presents OLS regression results of the firm-credit channel tests by exploiting the heterogeneous responses in the low income group between residents living in counties with high and low branch density. We utilize the pre-shock branch density data in 1994 and first totalize the county-level number of branches before the deregulation. We then scale this raw density by county total population. We split the low income group sample into high county branch density and low county branch density based on the county density median. Depression Index ranges from 0 (least depressed) to 21 (most depressed) and measures the severity of depression symptoms. D-WorkHours is a dummy variable that equals one if the respondent participates in the job market and has positive working hours in a given year, and zero otherwise. Log(WorkHours) denotes natural logarithm of one plus total number of hours worked in a given year. Log(EmpWeeks) denotes the natural logarithm of one plus total employment weeks in a given year. Log(TotalIncome) is the natural logarithm of one plus the respondent’s total income (in 2010 dollars), including wages, salaries, and income from farm or business. The key independent variable is IBBEA Index, ranging from 0 (most restrictive) to 4 (most deregulated). State controls include log of GSP per capita and log of total population. We include individual and gender-race-year fixed effects, as well as state-level linear time trends in all columns. All regressions use the sample weights provided by NLSY79. Heteroskedasticity-robust standard errors clustered at the state level are in parentheses. We test the coefficient difference between the high and low branch density counties and p-value is reported in the last row. *, **, and *** indicate significance at 10%, 5%, and 1%.

Depression Index D-WorkHours Log(WorkHours) Log(EmpWeeks) Log(TotalIncome)

High Density

Low Density

High Density

Low Density

High Density

Low Density

High Density

Low Density

High Density

Low Density

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) IBBEA Index -0.109 -1.077** -0.00389 0.0327*** -0.0280 0.237*** -0.0117 0.113*** 0.00562 0.225**

(-0.50) (-2.87) (-0.69) (4.55) (-0.69) (4.33) (-0.55) (3.73) (0.08) (2.83)

Log(GSP Per Capita) 0.430 1.623 0.213 0.137 1.953 1.621* 1.232 0.727 0.607 2.629**

(0.11) (0.54) (0.87) (0.96) (1.16) (1.74) (1.41) (1.60) (0.22) (2.02)

Log(Population) -1.346 -1.214 0.0718 -0.0225 0.569 -0.540 0.132 -0.287 -1.495 -0.916* (-1.46) (-1.03) (0.70) (-0.59) (0.74) (-1.59) (0.40) (-1.57) (-1.09) (-2.00) Individual FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Gender-Race-Year FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes StateLinearTrends Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Observation 3384 3149 8260 7651 8260 7651 8351 7715 7916 7330 R-Squared 0.412 0.343 0.514 0.464 0.559 0.499 0.549 0.486 0.564 0.520 P-Value 0.048 0.000 0.000 0.000 0.011

41

Table 9. Bank Branching Deregulation and Leverage

This table reports OLS regression results of the effect of the interstate bank branching deregulation on leverage condition in the low income group over 1992 to 2005. Leverage is the ratio of total liabilities excluding mortgages on residential property (e.g., debts on farm/business/other property, debts on vehicles etc.) over total assets excluding market value of residential property (e.g., total amount of money assets like savings, market value of farm/business/other property, market value of all vehicles and all other assets worth more than $500 etc.). D/I Ratio is the ratio of total liabilities excluding mortgages on residential property (e.g., debts on farm/business/other property, debts on vehicles etc.) over respondent’s total income (including wages, salaries, and income from farm or business). The key independent variable is IBBEA Index, ranging from 0 (most restrictive) to 4 (most deregulated). State controls include log of GSP per capita and log of total population. We include individual and gender-race-year fixed effects, as well as state-level linear time trends in all columns. All regressions use the sample weights provided by NLSY79. Heteroskedasticity-robust standard errors clustered at the state level are in parentheses. We test the coefficient difference between the high and low group and p-value is reported in the last row. *, **, and *** indicate significance at 10%, 5%, and 1%.

Leverage D/I Ratio Leverage D/I Ratio

High EFD

Low EFD

High EFD

Low EFD

High Density

Low Density

High Density

Low Density

(1) (2) (3) (4) (5) (6) (7) (8) IBBEA Index 0.0602 -0.0445 -0.0291 -0.0175 -0.0121 0.0191 -0.0145 -0.0398

(1.18) (-1.04) (-0.51) (-0.68) (-0.42) (0.38) (-0.39) (-1.09)

Log(GSP Per Capita) -0.984 -1.187 -0.0223 -2.560 -0.854 -1.402 -2.721* 0.331

(-0.84) (-0.86) (-0.02) (-1.67) (-0.70) (-1.33) (-1.80) (0.32)

Log(Population) -0.195 -0.409 -0.268 -0.551 0.200 -0.153 -0.482 -0.715 (-0.71) (-0.91) (-0.88) (-0.70) (1.24) (-0.80) (-1.08) (-1.60) Individual FE Yes Yes Yes Yes Yes Yes Yes Yes Gender-Race-Year FE Yes Yes Yes Yes Yes Yes Yes Yes StateLinearTrends Yes Yes Yes Yes Yes Yes Yes Yes Observation 5589 6511 4769 5528 6292 5718 5384 4853 R-Squared 0.255 0.356 0.254 0.286 0.350 0.258 0.278 0.259 P-Value 0.155 0.859 0.506 0.687

42

Appendix 1: Variable Definition Variable Definition Source State-Level Bank Branching Deregulation

IBBEA Index

We follow RS (2010) and contrast the index ranging from 0 (most restrictive) to 4 (most deregulated). We add one point to the index if the state (1) does not impose a minimum age restriction for acquisition, (2) allows de novo interstate branching, (3) permits interstate branching by acquiring a single branch; (4) sets the deposit-cap no less than 30%.

Rice and Strahan (2010)

CES-Depression Scale Depression Index A total score of seven items measuring symptoms of depression, ranging

from 0 (least depressed) to 21 (most depressed). NLSY79

D-Depression Index Above 8 A dummy variable that equals one if Depression Index is above the cutoff 8, and zero otherwise. NLSY79

D-Depression Index Above 10 A dummy variable that equals one if Depression Index is above the cutoff 10, and zero otherwise. NLSY79

D-Depression Index Above 12 A dummy variable that equals one if Depression Index is above the cutoff 12, and zero otherwise. NLSY79

% of Answers “0” Percentage of choosing 0 (Rarely/None of the time/1 Day) as the ratings among seven statements measuring symptoms of depression. NLSY79

% of Answers “1” Percentage of choosing 1 (Some/A little of the time/1-2 Days) as the ratings among seven statements measuring symptoms of depression. NLSY79

% of Answers “2” Percentage of choosing 2 (Occasionally/Moderate amount of the time/3-4 Days) as the ratings among seven statements measuring symptoms of depression.

NLSY79

% of Answers “3” Percentage of choosing 3 (Most/All of the times/5-7 Days) as the ratings among seven statements measuring symptoms of depression. NLSY79

CES-Depression 7 Dimensions

For each statement below, respondents choose from the following four alternatives as their ratings for the feelings over the past week: (1) Rarely/None of the time/1 Day (respondents choose 0 as the answer), (2) Some/A little of the time/1-2 Days (respondents choose 1 as the answer), (3) Occasionally/Moderate amount of the time/3-4 Days (respondents choose 2 as the answer), (4) Most/All of the times/5-7 Days (respondents choose 3 as the answer).

Trouble Eating I did not feel like eating; my appetite was poor. NLSY79 Trouble Concentrating I had trouble keeping my mind on what I was doing. NLSY79 Feel Depressed I feel depressed. NLSY79 Everything an Effort I felt that everything I did was an effort. NLSY79 Trouble sleeping My sleep was restless. NLSY79 Feel Sad I felt sad. NLSY79 Lethargic I could not get “going”. NLSY79 Labor Market Outcomes

D-WorkHours A dummy variable that equals one if the respondent participates in the job market and has positive working hours in a given year, and zero otherwise. NLSY79

Log(WorkHours) Natural logarithm of one plus total number of hours worked in a given year. NLSY79

Log(EmpWeeks) Natural logarithm of one plus total employment weeks in a given year. NLSY79

43

Log(TotalIncome) Natural logarithm of one plus the respondent’s total income (in 2010 dollars), including wages, salaries, and income from farm or business. NLSY79

Leverage Ratio

Leverage

The ratio of total liabilities excluding mortgages on residential property (e.g., debts on farm/business/other property, debts on vehicles etc.) over total assets excluding market value of residential property (e.g., total amount of money assets like savings, market value of farm/business/other property, market value of all vehicles and all other assets worth more than $500 etc.).

NLSY79

D/I Ratio

The ratio of total liabilities excluding mortgages on residential property (e.g., debts on farm/business/other property, debts on vehicles etc.) over respondent’s total income (including wages, salaries, and income from farm or business).

NLSY79

Individual-Level Characteristics

Education Highest grade/years of schooling completed by the respondent, ranging from 0 (never attend school) to 20 (8th year college or more). NLSY79

Age The difference between the birth year and the survey year. NLSY79

FamilyIncome7981 Natural logarithm of one plus the average total family income (in 2010 dollars) over 1979 to 1981. NLSY79

MotherEducation Maximum years of schooling received by the respondent's mother. NLSY79 FatherEducation Maximum years of schooling received by the respondent's father. NLSY79

D-BothParents A dummy variable that equals one if the respondent lived with both parents at the age 14. NLSY79

AFQTAbilityTest

Normalized percentile score after controlling for age groups among all the respondents in the 1980 Armed Forces Qualification Test (AFQT), which covers arithmetic reasoning, work knowledge, paragraph comprehension and numerical operations. Scores for the four sections were summed to form an aggregate AFQT score. A higher percentile indicates better learning aptitude.

NLSY79

RosenbergSelfEsteem An aggregate score ranging from 0 to 30 in the 1987 survey to measure the respondent's self-evaluation of personal worth. A higher value indicates a higher level of self-esteem.

NLSY79

PearlinSenseOfControl An aggregate score ranging from 7 to 28 in the 1992 survey to measure the respondent's sense of control over one's life. A higher value indicates a more positive self-concept and a stronger internal control over one's destiny.

NLSY79

D-RiskAttitude A dummy variable that equals one if the respondent answers “Yes” facing the question “Would you take job that could either double family income or cut income by a third” in the 1993 survey, and zero otherwise.

NLSY79

State-Level Controls Log(GSP Per Capita) Natural logarithm of state-level GSP per capita (in 2010 dollars). BEA Log(Population) Natural logarithm of state-level total population. BEA Partitioning Variables Individual-Level Partitioning Variable

Income Group

We first average total income (in 2010 dollars), including wages, salaries and income from farm or business, for each respondent over a three-year window (1989-1991) just before the sample period (1992-2005). We then divide the whole sample into high and low groups based on the state median income.

NLSY79

44

County-Level Partitioning Variables

External Finance Dependence

Following the literature (Rajan and Zingales, 1998; Duchin, Ozbas and Sensoy, 2010), we define EFD as the ratio of capital expenditures minus net cash flows from operating activities over the capital expenditures. We first aggregate firm-level values to the 3-digit SIC industry-level and calculate annual EFD for each industry. We next take the median of industry-year EFD over the long window (1980-2000) as the measurement of industry-level EFD. To arrive at the county-level EFD, we compute the employment-weighted EFD across all 3-digit SIC industries for each county-year. Then we average the county-year EFD over the pre-shock period 1989-1991 to gauge a persistent measurement of county-level EFD. We split the low income group sample into low county EFD and high county EFD based on the county EFD median.

Compustat, CBP

Branch Density

We utilize the pre-shock branch density data in 1994 and first totalize the county-level number of branches before the deregulation. We then scale this raw density by county total population. We split the low income group sample into high county branch density and low county branch density based on the county density median.

FDIC

45

Online Appendix

Table OA1. Bank Branching Deregulation and Clinical Depression Severity

This table repeats table 3 and reports OLS regression results of the effect of the interstate bank branching deregulation on the choice distribution of the CES-Depression Scale in the high income group over 1992 to 2005. D-Depression Index Above 8, D-Depression Index Above 10, D-Depression Index Above 12 respectively represents a dummy variable that equals one if Depression Index is above the cutoff 8/10/12, and zero otherwise. This set of dummies can measure the likelihood of clinical major depression. CES-Depression Scale is made up of seven statements evaluating the level of depression symptoms over the last week and respondents choose from the following four alternatives as their ratings: (1) Rarely/None of the time/1 Day (respondents choose 0 as the answer), (2) Some/A little of the time/1-2 Days (respondents choose 1 as the answer), (3) Occasionally/Moderate amount of the time/3-4 Days (respondents choose 2 as the answer), (4) Most/All of the times/5-7 Days (respondents choose 3 as the answer). % of Answers “0,” % of Answers “1,” % of Answers “2” and % of Answers “3” respectively denotes the percentage of choosing 0/1/2/3 as the ratings among the seven statements. The key independent variable is IBBEA Index, ranging from 0 (most restrictive) to 4 (most deregulated). State controls include log of GSP per capita and log of total population. We include individual and gender-race-year fixed effects in all columns. All regressions use the sample weights provided by NLSY79. Heteroskedasticity-robust standard errors clustered at the state level are in parentheses. *, **, and *** indicate significance at 10%, 5%, and 1%.

High Income Group

D-Depression Index Above 8

D-Depression Index Above 10

D-Depression Index Above 12

% of Answers “0”

% of Answers “1”

% of Answers “2”

% of Answers “3”

(1) (2) (3) (4) (5) (6) (7) IBBEA Index 0.00203 -0.000217 -0.000505 -0.000183 -0.00321 0.00597** -0.00259

(0.34) (-0.05) (-0.17) (-0.04) (-0.95) (3.02) (-1.36)

Log(GSP Per Capita) -0.0647 0.0413 0.0903 -0.0341 -0.0222 0.0403 0.0160

(-0.54) (0.43) (1.51) (-0.26) (-0.19) (0.79) (0.42)

Log(Population) -0.00117 0.0115 -0.0239 -0.0308 0.0342 0.0138 -0.0173 (-0.05) (0.45) (-1.47) (-1.24) (1.15) (1.18) (-1.67) Individual FE Yes Yes Yes Yes Yes Yes Yes Gender-Race-Year FE Yes Yes Yes Yes Yes Yes Yes Observation 7687 7687 7687 7687 7687 7687 7687 R-Squared 0.216 0.150 0.0948 0.350 0.217 0.169 0.173

1

Table OA2. Bank Branching Deregulation and Seven Dimensions of CES-Depression Scale

This table repeats table 4 and reports OLS regression results of the effect of the interstate bank branching deregulation on seven dimensions of the CES-Depression Scale in the high income group over 1992 to 2005. CES-Depression Scale is made up of seven statements evaluating the level of depression symptoms over the last week and respondents choose from the following four alternatives as their ratings: (1) Rarely/None of the time/1 Day (respondents choose 0 as the answer), (2) Some/A little of the time/1-2 Days (respondents choose 1 as the answer), (3) Occasionally/Moderate amount of the time/3-4 Days (respondents choose 2 as the answer), (4) Most/All of the times/5-7 Days (respondents choose 3 as the answer). Trouble Eating is “I did not feel like eating; my appetite was poor”. Trouble Concentrating is “I had trouble keeping my mind on what I was doing”. Feel Depressed is “I feel depressed”. Everything an Effort is “I felt that everything I did was an effort”. Trouble sleeping is “My sleep was restless”. Feel Sad is “I felt sad”. Lethargic is “I could not get ‘going’”. The key independent variable is IBBEA Index, ranging from 0 (most restrictive) to 4 (most deregulated). State controls include log of GSP per capita and log of total population. We include individual and gender-race-year fixed effects in all columns. All regressions use the sample weights provided by NLSY79. Heteroskedasticity-robust standard errors clustered at the state level are in parentheses. *, **, and *** indicate significance at 10%, 5%, and 1%.

High Income Group

Trouble Eating

Trouble Concentrating

Feel Depressed

Everything an Effort

Trouble sleeping Feel Sad Lethargic

(1) (2) (3) (4) (5) (6) (7) IBBEA Index 0.0154* 0.00724 0.00501 -0.00891 0.00599 -0.00706 -0.0108

(1.68) (0.55) (0.48) (-0.57) (0.42) (-0.61) (-0.81)

Log(GSP Per Capita) 0.457** -0.0811 -0.0746 0.135 0.421 -0.236 0.124

(2.14) (-0.29) (-0.24) (0.40) (0.98) (-0.92) (0.40)

Log(Population) 0.0328 -0.00202 0.117* -0.0640 0.0563 0.0734 -0.143** (0.34) (-0.02) (1.86) (-0.70) (0.63) (1.33) (-2.08) Individual FE Yes Yes Yes Yes Yes Yes Yes Gender-Race-Year FE Yes Yes Yes Yes Yes Yes Yes Observation 7687 7687 7687 7687 7687 7687 7687 R-Squared 0.195 0.182 0.191 0.275 0.241 0.205 0.244

2

Table OA3. Spatial Regression Discontinuity Design-Robustness Checks

This table presents OLS regression results of the effect of the interstate bank branching deregulation on the CES-Depression Scale in the low income group over 1992 to 2005 under different distance cutoffs, to mitigate cross-state lending concern. We exploit a spatial regression discontinuity design and restrict the sample to the contiguous county-pairs across the state borders. We further drop county-pairs with very close travel distance to address cross-state lending issue. We measure the distance in miles using the travel distance between two contiguous counties. Depression Index ranges from 0 (least depressed) to 21 (most depressed) and measures the severity of depression symptoms. The key independent variable is IBBEA Index, ranging from 0 (most restrictive) to 4 (most deregulated). State controls include log of GSP per capita and log of total population. We include county-pair and gender-race-year fixed effects, as well as state-level linear time trends in all columns. Heteroskedasticity-robust standard errors clustered at the county-pair level are in parentheses. *, **, and *** indicate significance at 10%, 5%, and 1%.

Drop <10 Miles Drop <15 Miles Drop <20 Miles Drop <25 Miles Drop <30 Miles

Depression Index

Depression Index

Depression Index

Depression Index

Depression Index

(1) (2) (3) (4) (5) IBBEA Index -0.276** -0.284** -0.355** -0.332** -0.317**

(-2.61) (-2.68) (-2.39) (-2.21) (-2.04)

Log(GSP Per Capita) 0.106 -0.0258 -0.0622 -0.538 -0.298

(0.08) (-0.02) (-0.04) (-0.34) (-0.15)

Log(Population) -1.005*** -0.992*** -0.985*** -0.941*** -0.882** (-4.05) (-3.97) (-3.80) (-3.56) (-3.21) Individual FE Yes Yes Yes Yes Yes Gender-Race-Year FE Yes Yes Yes Yes Yes StateLinearTrends Yes Yes Yes Yes Yes Observation 4456 4268 4082 3839 3388 R-Squared 0.220 0.224 0.228 0.228 0.230